What keeps happening

Sophie

Experienced manager, seven years in the organisation. Serving notice.

Sophie knows why the customer in Lyon requires invoices in a particular format. She knows that the 2019 ERP migration left an anomaly in quarterly reports. She knows which supplier contact actually answers the phone — and which ignores emails unless their director is copied in. She also knows which wording gets the best AI answers for the complex contracts her team handles every week. None of this knowledge appears in a document.

Sophie has just accepted an offer elsewhere — at a direct competitor. Her notice period is running. HR does not include what she knows in recruitment costs.

This scenario is universal. With every departure, formal procedures survive. The why disappears. Along with it goes the competitive advantage you spent years building, often to the company that just hired your employee. How do you prevent the loss of institutional expertise? The answer lies in an architectural decision, not a question of loyalty.

Marc

Business development specialist, three years in the company. Three weeks’ notice.

Marc built the company’s growth engine from the start. Email nurturing sequences in Make. LinkedIn automations with intent-based qualification. Lead scoring tuned on 18 months of data. Refined AI prompts for content creation, each representing dozens of iterations. A/B test results, all annotated in a personal Notion workspace. All of it, built with company time and resources, lives in his personal accounts.

Marc joins a competing startup. He takes his systems. In six weeks, he rebuilds what your team will spend a year reconstructing — if it manages at all.

In the first month after he leaves, the Zapier workflows keep running. Nobody knows why they are configured that way. Two months later, customer acquisition costs have tripled. The new hire starts from scratch, without test history, prompts or automations. Marc’s new employer has everything. What you paid to build has become your competitive disadvantage.

Reducing dependency on contractors and key experts is not about retention at any cost. It requires portable systems : when a business development specialist leaves, what they built stays. This is not an idealistic dream — it is an organisational decision. See the related challenges.

Claire

VP of Finance. Early AI adopter. Still in post.

Claire gets things done. Since January 2024, she has used Claude through her personal account to structure quarterly board presentations, model scenarios for an acquisition in progress, analyse the competitive position of the company’s three largest customers and prepare forecasts for the 2025–2027 strategic plan. The data she submits includes detailed margins by business unit, confidential revenue forecasts, acquisition targets and terms, and parts of the strategic plan that have not been made public.

Claire is not malicious. She is productive. She does not know that, on some plans, conversations may be used to improve models. She does not know that if she changes jobs, the account stays with her. Her company knows nothing about any of it.

The first time the question comes up — from an auditor, an investor conducting due diligence or legal counsel preparing a data governance report — the organisation answers: “We don’t know.” That “we don’t know” can cost tens of millions. It can also cost investor trust, the closing of a funding round or the company’s GDPR compliance.

Your legal and finance teams use ungoverned AI tools with sensitive data — as happens in 83% of organisations. The good news is that this is the most immediately addressable problem. It does not require a technical overhaul. It requires a decision and a framework. See the AI governance FAQ.